Research map / Image generation / Editing and style transfer
ChartEditor: A Reinforcement Learning Framework for Robust Chart Editing
Liangyu Chen, Yichen Xu, Jianzhe Ma, Yuqi Liu, Donglu Yang, Liang Zhang, Zihao Yue, Wenxuan Wang, Qin Jin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add both Chart-to-Code and editing SFT before RL (ablation) | ChartEditor · ChartEditVista (ablation) | average score: 18.5 -> 58.1 | Improved |
| add Chart-to-Code SFT to editing-only SFT (ablation) | ChartEditor · ChartEditVista (ablation) | average score: 53.2 -> 58.1 | Improved |
| add editing SFT to Chart-to-Code-only SFT (ablation) | ChartEditor · ChartEditVista (ablation) | average score: 54.3 -> 58.1 | Improved |
| add curriculum learning to RL (ablation) | ChartEditor · ChartEditVista (ablation) | average score: 56.6 -> 58.1 | Improved |
| add curriculum learning to RL (ablation) | ChartEditor · ChartEditVista (ablation) | execution rate: 73.6 -> 76.8 | Improved |
| add curriculum learning to RL (ablation) | ChartEditor · ChartEditVista (ablation) | text metric T_R: 53.0 -> 52.9 | No clear effect |
| add execution reward to format reward (ablation) | ChartEditor · ChartEditVista (ablation) | average score: 50.8 -> 54.6 | Improved |
| add rendering reward in place of format reward (ablation) | ChartEditor · ChartEditVista (ablation) | average score: 54.6 -> 57.2 | Improved |
| add format reward to rendering and execution rewards (ablation) | ChartEditor · ChartEditVista (ablation) | average score: 57.2 -> 58.1 | Improved |
| apply rendering reward RL to chart-to-code | Qwen2.5-VL-3B · ChartMimic Direct Mimic | overall score: 26.1 -> 54.5 | Improved |
| train with SFT plus RL using rendering reward | Qwen2.5-VL-3B · ChartEditVista | average score: 9.5 -> 58.1 | Improved |
| train with SFT plus RL using rendering reward | Qwen2.5-VL-3B · ChartEdit w/o Code (out-of-domain) | chart score: 24.3 -> 55.3 | Improved |
Closest papers by meaning
- ChartE$^{3}$: A Comprehensive Benchmark for End-to-End Chart Editing (ICML 2026 · similarity 0.94)
- ChartNet: A Million-Scale, High-Quality Multimodal Dataset for Robust Chart Understanding (CVPR 2026 · similarity 0.87)
- Breaking the SFT Plateau: Multimodal Structured Reinforcement Learning for Chart-to-Code Generation (ICLR 2026 · similarity 0.87)
- SketchVL: Policy Optimization via Fine-Grained Credit Assignment for Chart Understanding and More (CVPR 2026 · similarity 0.86)
- ChartSketcher: Reasoning with Multimodal Feedback and Reflection for Chart Understanding (NeurIPS 2025 · similarity 0.86)
- VCG-Bench: Towards A Unified Visual-Centric Benchmark for Structured Generation and Editing (ICML 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
